Master’s Thesis
“Do I like the product and you, because you have my size and color too?”
A study on the effect of similarity in skin color and body size on female consumer responses
Tamara Esmée Hartman
“Do I like the product and you, because you have my size and color too?”
A study on the effect of similarity in skin color and body size on female consumer responses
Master’s Thesis
Student name: Tamara Esmée Hartman Student number: S2069938
Study: M-COM Marketing Communication & Design Supervisor: Dr. J.J. van Hoof
Second Assessor: Dr. I. van Ooijen Date: 16 September 2019
2
Acknowledgments
First of all, I would like to thank my supervisor, Joris van Hoof. On the one hand, he gave me the feedback and supported that I needed, and on the other hand Van Hoof gave me a lot of freedom during the process, for example, in creating my own stimulus material. Also, I would like to thank my second supervisor Iris van Ooijen. She especially challenged me to get the most out of the statistical results by contributing her knowledge and ideas.
I would like to acknowledge professor Harry van de Kaap. He really took the time to give me advice considering my rather complicated statistical results. Harry made me see the light at the end of the tunnel again.
To all my models I would like to say, thank you for doing the photoshoot with me. I am not a professional photographer, but together we made this work! Keep celebrating diversity. Also, a big thanks to all the women who participated in the experimental survey.
In addition, I would like to thank my parents for supporting me during the process. You two always believed, and I love you both of you for that. Also, I want to thank my close uni-friends Alina and Cathlijn, you two have both been helpful and a fun distraction in this period. Finally, I would like to thank the one who was always there for me on the brightest and the worst days during this process. Sybren, you are the best.
Tamara Hartman September 2019
3
Abstract
Aim. This study investigates whether similarity in body size and skin color between advertised models and consumers increases positive consumer responses by presenting similar and dissimilar models in mascara advertisements. This study is performed since academic research focusing on positive consumer responses, as a result of similarity between individuals, is lacking. Also, previous studies considering the similarity-attraction effect mainly focused on psychological traits and not physical characteristics. An investigation into consumer effects because of similarity is relevant for theory since the similarity-attraction effect is argued to be one of the best-proven effects in social psychology. From a practical point of view, this study has societal relevance since the effect of portraying diverse models is investigated. Also, results can be used as input for marketing campaigns.
Method. An experimental 2 [Skin color: similar vs. dissimilar] x 3 [Body size: dissimilar body size below average BMI vs. similar body size average BMI vs. dissimilar body size above BMI]
between-subject design was conducted for the positive consumer outcomes, related to attitudes and purchase intention. Also, the moderating role of ethnic identity for skin color, and body satisfaction as moderator for body size was studied. In total, 157 women with an average BMI and white skin color participated in the online questionnaire.
Results. Analyses of respondents presented significant interaction effects for similarity in skin color and body size. The results indicated that presenting a similar body-sized model significantly increased the attitude towards the product and the model, but only when the skin color of the model was similar to the consumer, and thus white. Positive effects for the same variables were also found when presenting a dissimilar body-sized model with a BMI above average with white skin color. For body size, the moderator skin color significantly increased attitude towards the advertisement, but only for the medium body satisfied women. Also, ethnic identity was no moderator between skin color and positive consumer outcomes.
Conclusion. No main effects were found for skin color and body. However, the interaction effect indicated that a similar skin color is a prerequisite for similarity in body size to have a significant effect on attitudinal outcomes. Also, positive outcome effects were found when presenting dissimilar plus-size models with a similar skin color. Therefore, these two types of models are advised when advertising for white women with an average BMI.
Keywords: Similarity, Similarity-Attraction, Skin color, Body Size, Models, Ethnic Identity, Body Satisfaction, Attitudes, Purchase Intention.
4 Table of content
Acknowledgments ... 2
Abstract ... 3
Introduction ... 6
Theoretical framework ... 9
Judging physical appearance ... 9
Homophily ... 10
The similarity-attraction effect and perspectives ... 11
Reinforcement model ... 11
Information processing perspective ... 12
Similarity and skin color ... 13
Similarity and body size ... 13
Outcome effects of similarity in advertisements ... 14
Skin color similarity advertisements ... 15
Body size similarity advertisements ... 16
Interaction effect body size and skin color ... 17
Ethnic identification as moderator for skin color ... 18
Subquestion: Moderator body satisfaction for body size ... 19
Methods ... 20
Research Design ... 20
Stimulus material ... 21
Pre-test design ... 22
Results pre-test ... 23
Conclusion pre-test ... 25
Procedure experiment ... 25
Operationalization and measures experiment ... 26
Participants... 30
Descriptive statistics ... 31
Results ... 33
5
Hypothesis testing ... 33
Hypothesis 1: The effect of similarity in skin color ... 33
Hypothesis 2: The effect of similarity in body size ... 35
Hypothesis 3: Interaction effect between skin color and body size ... 36
Hypothesis 4: moderator ethnic identity for skin color ... 39
Subquestion: The moderator body satisfaction for body size ... 42
Discussion ... 46
Practical implications ... 49
Limitations and future research ... 50
Conclusion ... 53
Appendix I ... 67
Appendix II ... 70
Appendix III ... 73
Appendix IV ... 76
6
Introduction
Nowadays, more diverse models are displayed in advertisements compared to a few decades ago when skinny and white was ‘the rule’ (Baker, Sivyer & Towell, 1997).
According to the Runway Diversity Report, the Fashion Week runways of fall 2019 were more racially diverse than ever with an average of 39% female models of color. Also, there is a substantial increase in the representation of plus-size models during fashion week, from 14 plus-size models in spring 2016 to 50 plus-size models in fall 2018 (Tai, 2019). Currently leading fashion and cosmetic brands such as ASOS, Dove, and Wehkamp also display models with different body sizes and of different colors in their marketing campaigns (Howland, 2018; Unilever, n.d.; Mirck, 2019).
Besides the mere desirability of portraying a more diverse, and thus representative reflection of the population, the diversity approach may also have a positive effect in terms of brand and product responses. Especially because more diversity means more similarity between the model and consumer for a larger group in society. In contrast, white and skinny people might not recognize themselves anymore in these diverse advertisements, and this might decrease consumer outcomes since literature states that similarity increases attraction (Montoya, Horton & Kirchner, 2008). However, investigations into consumer responses because of similarity in physical characteristics lack, and a gap is addressed in literature (Montoya et al., 2008; Stevens, Owens & Schaefer, 1990; Perrier, 2008).
The effect that similarity is found to increase attraction is derived from homophily, whereby connections happen more often between similar individuals than between dissimilar individuals (McPherson, Smith-Lovin & Cook, 2001). This effect is argued to be one of the best-proven effects found in social psychology (Ajzen, 1974; Byrne, 1971;
Kaplan &, Anderson 1973; Layton & Insko, 1974; Montoya et al., 2008). Previous studies of the similarity-attraction effect mainly focused on psychological traits, with variables
7 such as attitudes and intelligence, and demographical characteristics such as age and education (McPherson et al., 2008).
The consumer effects of similarity in physical characteristics are both unexposed in the practical field and in literature. In the practical field, more and more diversity is presented, but it is unclear what the effects are and whether they have to do with a feeling of similarity between the advertised model and consumer. In theory, some explanations why skin color and body size might increase positive consumer responses exist, although solid social psychological investigations are lacking. For skin color, it is found that similar ethnicity is the most important factor for a connection between individuals when compared to age, religion, education, profession, and gender (McPherson et al., 2001). In particular, skin color can be seen as a physical characteristic of ethnicity, and therefore similarity in skin color might not only result in attraction but also have positive consumer outcomes (Gergen, 1967). Previous studies found an increase in positive consumer responses when similarity in skin color was detected, although positive results were mainly found among black participants (Appiah, 2001; Whittler & Spira, 2002). It remains unclear whether similarity in skin color has positive consumer effects on people with other skin colors.
In sociobiological studies, it is found that couples are more similar in body characteristics such as BMI (Silventoinen et al., 2003), length (Ginsburg, Livshits, Yakovenko & Kobyliansky, 1998), and obesity (Speakman, Djafarian, Stewart & Jackson, 2007) when compared to random individuals. However, in these studies, it is not investigated if this connection happens because people are more attracted to each other and thus, outcome effects of similarity in body size lack in social psychological literature (Perrier, 2008). Nevertheless, positive effects of portraying diverse body sizes are found in the practical field, especially for beauty and fashion brands such as Dove, ASOS and Wehkamp (Howland, 2018; Unilever, n.d.; Mirck, 2019).
8 To conclude, this study focusses on the positive consumer effects of similarity in body size and skin color, and the effects of similarity will be tested by randomly presenting both similar and dissimilar models in experimental advertisements. The theoretical goal of this study is to fill the existing gap in similarity research considering physical characteristics and consumer effects in a marketing context. Portraying diverse models can be socially relevant and interesting for marketing purposes, and therefore a practical goal is to give clarity considering the use of diversity and similarity as input for advertisements. Finally, an interaction effect between body size and skin color is investigated, which is entirely new in this field of study. For body size, the moderator body satisfaction is investigated, and for skin color, ethnic identity is studied as moderator. As a result of the previous explanations, the following research question is addressed:
Research question: “To what extent does similarity in body size and skin color between the model and consumer positively influence female consumer responses?”
9
Theoretical framework
In this theoretical framework, the similarity-attraction effect and consumer responses are discussed. At first, an introduction to judging individuals is given by using the in-group and out-group theory. Second, the concept of homophily is discussed. Third, the similarity- attraction effect which has roots in homophily is explained, including the two main perspectives. After, a substantiation is given for the incorporation of the physical characteristics: skin color and body size. Furthermore, the consumer responses of similarity advertisement are deliberated, from a general point of view to a specification for skin color and body size. At the end of the chapter, the moderators and hypotheses are introduced.
Judging physical appearance
In order to create understanding, humans categorize information which can be seen as a form of judging (Giles & Coupland, 1991). In the process of judging, people generally evaluate whether another individual is part of their in-group or out-group (Sumner, 1906).
The in-group formation divides individuals into two categories: individuals that are acknowledged similar, and those that are not. Nevertheless, this process of categorizing can differ from person and context (Allport, 1954). Initially, it was believed that individuals who had positive feelings towards the ingroup automatically also had negative feelings towards outgroups. However, it is found that positivity towards ingroup members is not related to negativity towards outgroup members (Allport, 1954; Brewer, 1999).
Furthermore, people are often judged based on their physical appearance since this is the first information that is directly available in a real-life interaction (Zebrowitz &
Montepare, 2008). According to Gergen (1967), individuals categorize others based on their physical differences and give these categories specific connotative meanings. This
10 process is ‘stereotyping’ when it takes place on a denotative level (Gergen, 1967). It is found that the appearance of others can even directly influence an individual’s behavior (Snyder, Tanke & Berscheid, 1977). Moreover, a favorable judgment towards another individual can elicit interpersonal attraction (Aron & Lewandowski, 2011). According to Goddard (2012), an interpersonal attraction is predicted by similarity, which means that because of similarity, interpersonal attraction can occur. For instance, it is found that couples with similar characteristics are attracted to each other (Goddard, 2012).
Homophily
In social psychology, the relationship between similarity and connection among individuals is called ‘homophily’ (McPherson et al., 2001). Homophily is an old concept that derives from the Ancient Greek ‘homou philia’. Initially, homophily was studied in relation to psychological traits and demographic characteristics whereby the concept was divided into status homophily and value homophily (McPherson et al., 2001; Lazarsfeld and Merton, 1954). Status homophily includes sociodemographic characteristics, such as race, sex, and ethnicity, whereas value homophily consists of beliefs, values, and attributes.
For example, it was found that similar values are essential for a vital relationship (Richardson, 1940). However, lacking in the explanation of Lazarsfeld and Merton (1954) is similarity-based on physical characteristics. In more recent homophily publications, attractions based on physical appearance is also not mentioned (McPherson et al., 2001).
Moreover, social psychologist Newcomb (1961), found in an experiment that individuals with similar attitudes and beliefs liked each other more than individuals with differing attitudes and beliefs. In particular, romantic relationships are often used as a context when studying homophily (Morry, 2007). Especially anthropological studies relate to the concept of ‘homogamy,’ which means homophily in marriage (Lazarsfeld & Merton, 1954). Academic studies considering homogamy have also not included the effect of
11 similar physical appearance. Nevertheless, from a sociobiological perspective, literature states that connections based on similar physical characteristics occur. This literature includes positive assortative mating, a pattern whereby individuals with similar characteristics (phenotypes), and cultural traits, connect more often with each other than individuals with dissimilar phenotypes (Domingue et al., 2014; Speakman et al., 2007).
This concept contains genetic relatedness and concludes that individuals are genetically more related to spouses than to random individuals (Speakman et al., 2007). Although the concept is not often used for social-psychological studies, it gives a reason to believe that homophily does exist for physical characteristics.
The similarity-attraction effect and perspectives
The specific relationship between similarity and attraction can be explained as the similarity-attraction effect. Within this effect, which is derived from homophily, increased similarity towards a person is associated with an increased attraction (Montoya & Horton, 2012). The similarity-attraction effect is argued to be the best-proven relationship in behavioral science (Berger, 1973). A meta-analysis with more than 300 studies concludes that similarity produces a positive effect on attraction (Montoya, Horton & Kirchner, 2008). Moreover, the main explanations of the similarity-attraction effect are the reinforcement model and the information processing perspective.
Reinforcement model
According to the reinforcement model, people have a central need to view the world as coherent and logical (Byrne, Griffitt & Stefaniak, 1967). Within this view, a similar individual, for example, an individual who shares similar political views, is seen as reinforcing when compared to self. As a result, the person is evaluated positively, which leads to an attraction (Byrne et al., 1967). Furthermore, the reinforcements occur on an
12 unconscious level (Byrne, Rasche, & Kelley, 1974). Therefore, the model can be seen as an affective model. Although the similarity-attraction effect of this model is not fully proven in field studies (Montoya et al., 2008).
Information processing perspective
From the information processing perspective, attraction to another person is determined by the available positive information one has of another person (Ajzen, 1974; Kaplan &
Anderson, 1973; Tesser, 1971). According to this perspective, attraction is a function of the valence and weight of information that is received about an individual based on (dis)similarity (Montoya et al., 2012). Compared to the reinforcement model, the information processing perspective is a more cognitive approach. First of all, information about an attribute is assigned valence, for example when a person is told that another person is similar to us, the person will instantly like the other more because the person will assume that the other has positive attributes. This effect will occur because people generally evaluate their own attributes positive (Ajzen, 1974).
Second, the information about another person is weighted, this means that the more information one receives about a specific attribute of the other, the more important that specific attribute will be in determining attraction. Therefore, more information about attributes will produce more polarized judgments than less information (Montoya et al., 2012). Third, the salience of information is a determinant whereby the more attention an attribute receives, the more the information will affect one's judgment (Montoya et al., 2012). Thus, according to the information processing perspective attraction based on similarity depends on the valence, weight, and salience of information.
13
Similarity and skin color
Generally, skin color is used as a characteristic of race or ethnicity when judging individuals (Gergen, 1967). For instance, skin pigmentation was found to be an extremely crucial factor in the favorability of black applicants in work environments (Harrison &
Thomas, 2009). The result of this study showed that lighter black applicants were judged more positively than dark black applications and that skin color is more important for job hiring and recommendation than previous work experiences and educational accomplishments (Harrison & Thomas, 2009).
McPherson et al. (2001) found that a similar race and ethnicity are the most influential factors for a connection between individuals. For instance, in a study by Marsden (1987), individuals were asked to look back over the past six months and indicate whether they discussed important matters with individuals from other races. Only 8% of the individuals indicated to have done this (Marsden, 1987). In sociobiological studies, skin pigmentation on foreheads is associated with assortative mating, which means that couples are more alike in skin pigmentation than when compared to a random individual (Roberts & Kahlon, 1972). This finding indicates that individuals with similar skin color connect more often. However, positive assortative mating for skin color mainly depends on the cultural context (Parra, Kittles & Shrivers, 2004). Based on the similarity-attraction effect, it is likely to believe that similarity in skin color will cause attraction and lead to other positive outcome effects.
Similarity and body size
In the fashion industry, diversity in both skin color and body size were inconceivable until a few years ago (Tai, 2019). In advertisements, still mainly thin models are used to represent a product, and the reasoning behind this might be derived from the halo-effect,
14 which indicates ‘what is beautiful is good.’ Thin models were still seen as ideal and more beautiful compared to average and plus-size models in the western world in the year of 2004 (Halliwell & Ditmar, 2004). Nonetheless, thin models only represent a small group of the female population. Therefore, a similarity-attraction effect between consumer and models in advertisements might solitary be available for a relatively small group of thin women. Based on the reasoning of the similarity-attraction effect, a congruency in body size between the presented spokesperson, a person that presents the product or service in the advertisement, and the consumer will result in positive consumer responses (Montoya et al., 2008). Positive effects of similarity in body size are found before, for instance, an attraction based on similarity in body size is found to result in friendships (Crosnoe, Frank,
& Strassman-Mueller, 2008). Nevertheless, literature considering this topic is in short supply, especially in social psychological studies.
Outcome effects of similarity in advertisements
Similarity between the viewer of an advertisement and the spokesperson can lead to positive outcomes. For instance, literature shows that congruence between the multicultural background of the spokesperson and consumer increased the effectiveness of advertisements (Deshields, Kara, De Los Santos, 1999). Perceived age similarity between models and consumers was to increase interpersonal attraction and source credibility (Steinhaus & Lapitsky, 1986). Also, older consumers found older models more attractive and were more likely to purchase the items that were promoted (Kozar & Damhorst, 2008).
Positive attitudes towards the advertisement and the presented product are described in similarity-attraction literature as an outcome effect of similarity (Aaker, Brumbaugh & Grier, 2000). According to the American Marketing Association, attitude can be defined as an overall evaluation of an impression (AMA, 1995). Specifically, a positive attitude towards a message source is believed to be necessary for the effectiveness
15 of communication (Steinhaus & Lapitsky, 1986). To conclude, an increase in purchase intention because of similarity in skin color, sex roles, body size, and age are also found as outcome effects in advertisements (Kozar & Damhorst, 2008; Perrier, 2008; Webster, 1994; Whittler & Spira, 2002; Woodside & Davenport, 1974). Purchase intention is the strongest predictor of behavior and can be seen as an indication of the preparedness to behave (Fishbein & Ajzen, 1975).
Skin color similarity advertisements
In advertising studies, the importance of similarity in skin color for positive consumer responses is addressed. Literature shows that black respondents perceive themselves more similar to, and identify more strongly with, black spokesmen in advertisements compared to white spokesmen (Appiah, 2001). This effect was stronger when individuals had a strong connection with their ethnic identity (Appiah, 2001). Considering outcome effects, Whittler and Spira (2002), concluded that product and overall advertisement evaluations were more favorable when black respondents were presented with black models. However, this was only the case when the respondents identified with black culture.
Moreover, Appiah (2001) argued that attitudes towards the spokesperson were more positive when both the presenter and respondent were black. Furthermore, an increase in purchase behavior was also found as a result of ethnic identification with the spokesperson in the advertisement (Webster, 1994; Whittler & Spira, 2002). The studies mentioned here found positive consumer responses, although these studies focused on a black study group. Since the outcome effects for people with other skin colors are unclear, the following hypotheses are proposed:
16 H1. A higher similarity in skin color between the model and the consumer leads to a) a
more positive attitude towards the ad in general, b) a more positive attitude towards the product in the ad, c) a more positive attitude towards the model in the ad, and d) a higher purchase intention for the product in the ad.
The word ‘spokesperson’ is often used in literature when a model in an advertisement presents a specific product or service. Therefore the scale for the model in the ad, presented in Appendix I, is referred to as ‘attitude towards the spokesperson.’ The word ‘model’ is used for this study since spokesperson and model have the same intention, and the word is more suited in the context of this study.
Body size similarity advertisements
The specific relationship between similarity in body size and positive consumer responses is not extensively investigated in academic literature. Nevertheless, it is important to investigate similarity for body size, since the lack of research and need for information into this concept is addressed (Perrier, 2008; Montoya et al., 2008). Furthermore, marketing practices such as Dove’s ‘real beauty campaign’ and ASOS current campaign-style suggests that marketers can use more realistic and socially responsible representations of female body images that match the appearance of the female population to increase positive consumer responses. For instance, Dove supports diversity by displaying multi-colored and plus-size models in their real-beauty campaigns (Unilever, n.d.). According to corporate publications, Dove is the largest soap brand worldwide, with a higher turnover than all competitor soap brands together (Unilever, n.d.). Since the launch of the real beauty campaign sales increased from 2.5 billion in 2004 to 4 billion in 2014 (Neff, 2014).
Moreover, online fashion store ASOS promotes fashion on models that differ in color, size, length, and weight (Howland, 2018). After the adoption of this approach,
17 placed orders increased by 30% compared to the period of the previous year, according to a self-published report (Howland, 2018). Also, consumers responded very positively towards the new approach of the brand on social media (Howland, 2018). In January 2018, ASOS was the e-commerce leader in the United States, with 27% more sales than competitors (WGSN, 2018). The representation of different sized models means that more body sizes and figures of women are represented in advertisements. This assumingly leads to a similarity-attraction effect for a larger group of females than before. However, it is uncertain that the presentation of different sized models, and possibly a better similarity- attraction effect, is solely responsible for the successes of Dove and ASOS. As a result of the positive findings of the similarity-attraction effect, the general positive consumer responses to similarity mentioned at the beginning of this chapter, such as the increase of source credibility (Steinhaus & Lapitsky, 1986), and the practical examples of Dove and ASOS, the following hypotheses for body size are proposed:
H2. A higher similarity in body size between the model and the consumer leads to a) a more positive attitude towards the ad in general, b) a more positive attitude towards the product in the ad, c) a more positive attitude towards the model in the ad, and d) a higher purchase intention for the product in the ad.
Interaction effect body size and skin color
As described in hypotheses 1 and 2, main effects for body size and skin color are expected.
Moreover, an interaction effect between skin color and body size for the dependent variables is also expected. Similarity in skin color, as a physical element of ethnic identity, is found to be extremely important to increase the attraction between individuals (McPherson et al., 2001). Since skin color is an important element in order to find attraction, it is assumed that a similar skin color with the model is needed for body size to
18 have a stronger effect, and thus more positive consumer outcomes. As a result, the following hypotheses are stated:
H3. Interaction between body size and skin color exists and leads to a) a more positive attitude towards the ad in general, b) a more positive attitude towards the product in the ad, c) a more positive attitude towards the model in the ad, and d) a higher purchase intention for the product in the ad.
Ethnic identification as moderator for skin color
Ethnic identity can be defined as an individual’s membership in a specific social group combined with the value and affective elements given to that membership (Phinney, 1992).
In similarity studies considering skin color, ethnic identification, also explained by cultural identification or ingroup orientation, was in some cases found to be a moderator between similarity in skin color and consumer outcomes that increased outcome effects (Appiah, 2001; Avery, 2003). For example, it is found that black applicants with low out-group orientations, and a high ethnic identity connection, experienced a positive effect of black spokesman in the advertisement. In contrast, black people with higher out-group orientations, and lower connection to ethnic identity experienced the ads with only diversity content as negative (Avery, 2003). For white applicants, ad diversity attracts some white respondents while deterring others (Avery, 2003). Since previous studies found contradicting results among white people, ethnic identification is hypothesized as a moderator. As a result, the following moderator is proposed:
H4. A stronger connection to ethnic identity leads to a) a more positive attitude towards the ad in general, b) a more positive attitude towards the product in the ad, c) a more positive attitude towards the model in the ad, and d) a higher purchase intention for the product in the ad, when compared to a weak connection to ethnic identity.
19
Subquestion: Moderator body satisfaction for body size
In previous literature, the concept of body satisfaction, defined as a stable personality characteristic, is found to be a moderator of vulnerability towards images in the media (Posavac, Posavac & Posavac, 1998). Moreover, individuals who are satisfied with their bodies are more favorable towards more substantial sizes, compared to body-dissatisfied people (Willinge, Touyz & Charles, 2006). Based on the outcome that body-satisfied people are more open to larger sizes, it might be that the similarity-attraction effect is weaker among women with higher body satisfaction. Although, it is not entirely clear whether body satisfaction is a moderator here since literature in this field is lacking.
Therefore, the following subquestion is proposed:
Subquestion: “To what extent is body image satisfaction a moderator between body size and positive consumer responses ?”
To summarize, the theoretical framework provided the theoretical background for similarity found in academic studies. The similarity-attraction effect is explained, including the roots of the theory and the two main perspectives. Also, the importance of the physical characteristics in relation to similarity is presented, and the aim for investigating body size and skin color as important elements is addressed. Precious investigations of consumer effects as a result of similarity in advertisements are discussed, and both theoretical and practical findings are used as input for the hypotheses.
20
Methods
Within this method section, the research design, stimulus material, pre-test design, pre-test results, the operationalization of the main experiment and procedure are described.
Research Design
An experimental between-subject design is used for this investigation. The experiment tests whether participants who are presented with a similar model in body size and/or skin color will report higher outcomes compared participants presented with a dissimilar model in body size and/or skin color. Solely women are examined for this study since this is more practical for methodology purposes. A specific group of women with an average BMI and white skin color is studied in this experiment to measure possible effects in a reliable and valid manner. As a result, similar conditions will consist of ‘average BMI’ for body size and ‘white’ for skin color. Furthermore, the moderator's ethnic identity for skin color and body satisfaction for body size are included in the model. The research design is presented in Figure 1.
Figure 1: Experimental design
• Body size BMI below average (dissimilar)
• Average BMI (similar)
• Body size BMI above average
(dissimilar) Attitude towards
the model in the ad
Manipulations Independent variables Dependent variables
• White skin color (similar)
• Black skin color (dissimilar)
Actual body size
Actual skin color
Attitude towards product in the ad Attitude towards the ad in general
Ethnic identity
Purchase intention towards product
in the ad Body satisfaction
21
Stimulus material
In total, six manipulations are created combining the two conditions of skin color: white (similar) and black (dissimilar), and the three conditions of body sizes related to a BMI, with body size below the average 18,5 (dissimilar), body size with an average BMI between 18,5 and 24,9 (similar), and body size with a BMI above average of >25 (dissimilar). In order to be more confident that the effects will not dependent on the clothing and pose, the models in the advertisements are similar in clothes and modeling position.
Based on existing literature and findings in the media, a specific promotion product is chosen for the advertisement. First of all, a product is chosen instead of a service because consumer perceives more risk in buying a service than a product (Murray and Schlacter, 1990). Since hedonic products might result in affective reactions, they are preferred to the utilitarian product in this study, since affective reactions are needed in order to measure attraction (Chang, Chen & Tang, 2009). Product types that have a connection to body sizes or skin colors, such as a foundation or food, cannot be used in this study since these products are not applicable to all models.
A product that fulfills the needs of low-risk, hedonic, and universal in use is mascara. In the United States, 65% of the female respondents indicated to use mascara in 2018 (NHCS, n.d.). Furthermore, mascara is the most popular beauty article among Dutch women. Results of a study showed that 74% of the Dutch females used mascara (Nu.nl, 2010). Moreover, color preference is also not a problem with this product since mascara is generally black. As a result, the final advertisements consist of a total body picture of the model in order to measure the effects, and a picture of the face that focuses on the eyes is also shown since this is customary in mascara advertisements. To conclude, the final designs of the advertisements are presented in Figure 2, larger versions of the advertisements can be found in Appendix II.
22 Figure 2: Design of advertisements
Pre-test design
Before the relationship between the variables can be investigated, it is essential to be sure that all manipulations are understood correctly. For this reason, a manipulation check in the pre-test is performed to test whether respondents understood the body size and skin color of the models correctly. Moreover, the attractiveness and credibility of the model were also measured as an extra element. All questions were measured on a seven-point Likert scale ranging from strongly disagree (1) to strongly agree (7).
A pre-test was created in Qualtrics with the six manipulations and four subjects related to questions about the body size of the model (BMI above average vs. BMI below average), the skin color of the model (black skin color vs. white skin color) the attractiveness of the model (attractive in general, the face is attractive and the body is attractive) and the credibility of the model (is the model credible). The same questions were asked for every advertisement. The advertisements were shown randomly so that the order of the advertisement differed per person.
23
Results pre-test
In total, 21 women filled in the pre-test. At first, the respondents were asked if they knew the model. One respondent knew one of the models, and therefore, this respondent is removed from the analysis. As a result, 20 respondents are used for the pre-test analysis.
The main goal was to investigate whether respondents indicated the correct BMI and skin color for the model. The used scale for all questions was a seven-point Likert scale ranging from (1) strongly disagree to (7) strongly agree. The results of the pre-test are presented in Table 1, and an extensive explanation of the results per manipulation can be found in Appendix III.
24
25
Conclusion pre-test
Overall, the results in Table 1 and Appendix III indicated that the models with a BMI below average are seen to have a slimmer body size than average, the models with an average BMI are not seen as slimmer than average nor heavier than average, and the models with a BMI above average are seen as heavier than average. Moreover, the black models were seen to have a dark skin color, and the white models were seen to have a white skin color.
Therefore, all manipulations are understood correctly and can be used as advertisements.
For both black and white models, the attractiveness of body was highest with the manipulation that had an average body size. Moreover, for black models, the credibility was lowest for the model with a BMI above average, while for white models, the credibility was lowest for the model with a BMI below average.
Procedure experiment
The gathering of the data is completed via a survey in Qualtrics. At first, in the online survey, participants were given the option to fill in the survey in Dutch or English. Also, the introduction explicitly stated that only responses of women were needed. Further, it was explained that the participates were going to see an advertisement and had to answer questions related to the advertisement to measure consumer responses. The goal of the study was stated, and the approximate duration time of the survey was given. Moreover, the option to leave the survey at any given time is presented, and the possibility to win a gift card is mentioned. The first question in the survey is a gender question. For male respondents, the survey ends since only females are used in this study. The women that continue are asked to pay close attention to the ad before the advertisement is shown.
However, it is possible for the respondent to go back to the advertisement and have an extra look. After that, randomly, one of the six conditions is presented to the participant.
26 Next, the participant is asked if she uses mascara and which mascara brand she prefers to use. Furthermore, questions related to the dependent variables are asked. First, the attitude towards overall advertisement is measured, followed by the attitude towards the product in the advertisement and the model that presents the product. Purchase intention is the final measured dependent variable, and four questions measure the perceived similarity with the model. After these questions, participants are asked to give their weight in kilo’s and length in centimeters. This question is essential since BMI-score can be calculated via these measures. Also, the question is asked whether the respondent has a white or black skin color. In this state of the survey, all respondents continue the survey, independent of their body size and skin color.
The reason why body size is asked in the end is that it is assumed that the bounce rate with a sensitive question is lower when respondents have almost finished a survey, compared to when they start with the survey. For the same reason, the moderator's body satisfaction and ethnic identity are asked at the end of the survey. To finalize the data collection questions about the demographics, age, and education level are asked. Also, in the end, respondents can leave their e-mail address if they want to win the gift card or receive the results of the study.
Operationalization and measures experiment
The operationalization and measures for the variables in the main experiment are described in this part of the method section.
Body size
A formula to measure weight is the W/H2 ratio of Quetelet, nowadays called the Body Mass Index (Keys, Fidanza, Karvonen, Kimura & Taylor, 1972). Body Mass Index (BMI), can be calculated with the mathematical formula: Kg/m2. BMI covers more elements than,
27 for example, waist circumference and is suitable for self-measurement (Hartstichting, 2018 April 20th ). In this study, BMI is used to indicate the respondent's body size. Solely respondents with average body size are useful for the experiment. The scale defines a BMI below 18,5 underweight, between 18,5 and 24,9 average and above 25 obese (Hartstichting 2018 April 20th). In Appendix I, the items to indicate the body size of the respondents can be found.
Skin color
At the beginning of the survey, a question checks whether the participants have a white or black skin color. The specific question related to skin color can be found in Appendix I.
Attitude towards the ad in general
The scale ‘attitude towards the ad’ exists of six items measured on a seven-point Likert scale from strongly disagree to strongly agree, and measures a person’s reaction towards an ad (De Pelsmacker, Geuens & Anckaerts, 2002). An example item is ‘I got a very positive impression.’ The item ‘exaggerated’ is left out since the item does not fit in the context.
Moreover, the Cronbach’s alpha found for this construct is relatively high, with a score of .89. In Appendix I, all five scale items can be found.
Attitude towards the product in the ad
The scale ‘attitude towards the product in the ad’ exists of four items. A seven-point Likert scale from strongly disagree to strongly agree is used for this study since this is in line with the other dependent variables. The scale assesses the person’s attitude about the product featured in the advertisement. An example is an item ‘The product in the ad is attractive’
(Lepkowska-White, Brashear & Weinberger, 2003). The Cronbach’s alpha for the four- item construct is high with a score of .91. The scale can be found in Appendix I.
28 Attitude towards the spokesperson (model)
The scale ‘attitude towards the spokesperson (likeability)’ evaluates a person’s opinion of the model featured in the ad focusing on the spokesperson favorability. In this study, the spokesperson is referred to as the model in the advertisement. The scale developed by Whittler and Dimeo (1991) exists of four bi-polar adjectives measured via a seven-point semantic differential scale. An example bi-polar adjective is likeable/unlikeable. The item attractive/unattractive is added to the scale. A Cronbach’s alpha of .88 indicates high reliability for this scale. In Appendix I, the items of the scale can be found.
Purchase intention towards the product in the ad
This scale measures the likelihood of a person buying the product featured in the ad with three statements measured via a five-point Likert-scale. However, a seven-point Likert scale is used. An example statement is: ‘If I were looking for this type of product, my likelihood of purchasing the product in the ad would be high.’ The scale developed by Lepkowska-White, Brashear, and Weinberger (2003) is more specific than a general purchase intention scale.
A Cronbach’s alpha .91 is found for the construct, which indicates high reliability. Also, all the items of the construct can be found in Appendix I.
Perceived similarity with the spokesperson
The perceived similarity between the participant and the model can be controlled with items from the scale ‘attitude towards the spokesperson similarity’ (McKiran, Smith &
Hamayan, 1983). An example item of this scale is: ‘How similar are you to the model pictured in the ad on appearance.’ Solitary the items that focus on the physical characteristics of the spokesperson are used for this study. Therefore, items appearance and cultural background are derived from the scale and measured on a seven-point Likert scale from (1) not at all similar to (7) very similar. Moreover, two self-created items: ‘perceived similarity in body
29 size’ and ‘perceived similarity in skin color’ are added in order to measure the perceived similarity. As a result, a Cronbach’s alpha .79 is found, which indicates a reliable construct.
In Appendix I, the composed scale can be found.
The moderator ethnic identification
The scale ethnic identification measures with five, seven-point Likert-type, statements the degree to which a person expresses a certain attachment to a specific ethnic group and have a positive feeling towards it. The scale used is developed by Appiah (2001) and is based on the multigroup ethnic identity measure (Phinney, 1992). An example item of ethnic identity is:
‘I feel a strong attachment to my ethnic group.’ Furthermore, a Cronbach’s alpha of .87 is found, which indicates high reliability. In Appendix, I, the items for the scale can be found.
The moderator body satisfaction
The body areas satisfaction scale (BASS) is a subscale of the multidimensional body-self relations questionnaire and is frequently used to asses the evaluation of individuals’ body image (Giovannelli, Cash, Henson & Engle, 2008). The body satisfaction scale consists of 9 items asked via a five-point Likert scale from very dissatisfied to very satisfied, but a seven-point Likert scale is used in this study (Giovannelli et al., 2008). An example item of this scale is: ‘Body satisfaction towards weight.’ A Cronbach’s alpha of .84 is found for the construct. The specifications of the torso items are explained in a study by Hrabosky et al., (2009). In Appendix I, the scale items are presented.
30
Participants
In total, 348 female respondents participated in the experiment. However, 65 respondents were removed from the dataset since the answers were incomplete. Still, 283 women filled in the survey correctly. After removing 3 respondents that recognized the presented models, 280 women remained. Since the study focusses on white women with an average BMI, 32 black women and 65 women with a BMI below or above average were removed from the analyses, resulting in 183 remaining women for the analyses. Moreover, 25 women were excluded from the analyses because these women indicated not to wear mascara. As a result, the remaining 158 mascara-using women with white skin color and average BMI are investigated for the assumptions of the study.
Since differences between groups will be investigated for multiple dependent variables, factorial ANOVA analyses are performed. In order to check for the assumptions of normality and outliers, standardized residuals for all dependent variables and moderators are checked. Multivariate outliers below -3 or above +3 indicate that something unusual is happening and therefore, these items have to be removed (Rossi, 2010). With a score of 3.15, one outlier was removed, for the dependent variable purchase intention. As a result, 157 respondents were used for the ANOVA analyses. In Table 2, the distribution of the participants per condition is presented.
Table 2: Distribution of participant per condition
Conditions N
1. Dissimilar skin color and dissimilar body size with BMI above average 30
2. Dissimilar skin color and similar body size with average BMI 33
3. Dissimilar skin color and dissimilar body size with BMI below average 28
4. Similar skin color and dissimilar body size with BMI above average 25
5. Similar skin color and similar body size with average BMI 21
6. Similar skin color and dissimilar body size with BMI below average 20
Total 157
31
Descriptive statistics
In total, 157 respondents, consisting of mascara-using women with an average BMI and white skin color, were used for the data analyses. Considering the descriptive statistics, both age and education level are questioned in the survey and can be found in Table 3.
Table 3: Characteristics of respondents
M SD N %
Age 25.5 7.3 157 100%
Education level category*
Low education level 15 10%
Medium education level 21 13%
High education level 121 77%
*The education level categories are explained in Appendix IV
The randomization of education per condition is tested with a chi-square test, and the results indicate that the percentage of participants within the conditions did not differ by education level χ² (10, N = 157) = 6.70, p = .75. The percentage per conditions is presented in Table 4.
Table 4: Randomization of education per condition
Condition
Level of education
Low education Medium education High education Total %
Condition 1 13.3% 6.7% 80% 100%
Condition 2 9.1% 9.1% 81.8% 100%
Condition 3 14.2% 17.9% 67.9% 100%
Condition 4 8% 12% 80% 100%
Condition 5 4.8% 14.2% 81% 100%
Condition 6 5% 25% 70% 100%
32 The randomization of age per condition is tested with a chi-square test, whereby age is divided into two categories based on the median score of 24. The results indicate that the percentage of participants within the conditions did not differ significantly by age category χ² (5, N = 157) = 4.70, p = .45. The percentage per conditions is presented in Table 5.
Table 5: Randomization of age category per condition
Condition
Age
Age category
below 24 Age category
above 24 Total %
Condition 1 73.3% 26.7% 100%
Condition 2 66.7% 33.3% 100%
Condition 3 64.3% 35.7% 100%
Condition 4 60% 40% 100%
Condition 5 57.1% 42.9% 100%
Condition 6 45% 55% 100%
33
Results
Hypothesis testing
In order to compare the main effects of the groups used in this study, multiple full-factorial ANOVA’s are conducted. First, the main effects of actual similarity in skin color were analyzed. Second, the main effects of body size were explored, and third possible interaction effects were tested. At the end of this chapter, the moderator ethnic identity for skin color and the moderator body satisfaction for body size were explored as fixed factors in ANOVA analyses.
Hypothesis 1: The effect of similarity in skin color
The actual similarity in skin color contains two levels: a similar condition (white skin color) and a dissimilar condition (black skin color). In total, 66 respondents were confronted with a similar colored model, and 91 respondents were confronted with a dissimilar colored model. The effects of similarity in skin color are tested for:
H1a: Attitude towards the advertisement in general;
H1b: Attitude towards the product in the advertisement;
H1c: Attitude towards the model in the advertisement;
H1d: Purchase intention for the product in the advertisement.
The outcomes are presented per dependent variable.
The main effect of attitudes towards the advertisement in general for actual similarity in skin color yielded an F-ratio of F(1,155) = .42, p = .52, indicating no significant differences between the groups who had seen a model with a similar skin color (M = 3.13, SD = 1.12) and a dissimilar skin color (M = 3.25, SD = 1.26).
34 The main effect of attitude towards the product in the advertisement for actual similarity in skin color yielded an F-ratio of F(1,155) = .16, p = .69, indicating no significant differences between the groups with a similar skin-colored model (M = 3.54, SD = 1.05) and a dissimilar skin-colored model (M = 3.62, SD = 1.26).
The main effect of attitude towards the model in the advertisement for actual similarity in skin color yielded an F-ratio of F(1,155) = 1.16, p = .21, again indicating no significant between the groups who had seen a model with a similar skin color (M = 4.26, SD = 1.09) and a model with a dissimilar skin color (M = 4.51, SD = 1.30).
The main effect of purchase intention for the product in the advertisement for actual similarity in skin color presented an F-ratio of F(1,155) = .12, p = .73, indicating no significant differences between the groups who had seen a similar skin-colored model (M
= 2.83, SD = 1.14) and a dissimilar skin-colored model (M= 2.90, SD = 1.39).
As a result of the outcomes, hypothesis 1a,1b,1c, and 1d are rejected. The overview of the rejected statistics can be found in Table 6.
Table 6: Summary of statistics skin color similarity
Independent Variable Dependent Variable F p
Skin color Overall score .64 .42
Attitude towards the ad in general .42 .52
Attitude towards the product in the ad .16 .69
Attitude towards the model in the ad 1.16 .21
Purchase intention .12 .73
35 Hypothesis 2: The effect of similarity in body size
The body size condition is divided between the similar body-sized model and the dissimilar body-sized model. However, the dissimilar body size model exists of two categories: the dissimilar model with a below-average BMI, and the dissimilar model above-average BMI.
The dissimilar conditions cannot be combined since the outcomes for the conditions are different. Considering the distribution, 48 participants judged an advertisement with a dissimilar model with a BMI below average, 55 participants judged an advertisement with a dissimilar model with a BMI above average, and 54 participants saw a model with a similar body size existing of an average BMI. Moreover, ANOVA analyses are performed for the 3 categories of body size and the dependent variables.
The main effect of attitudes towards the advertisement in general for actual similarity in body size yielded an F-ratio of F(2,154) = 1.83, p = .16, indicating no significant differences between the groups who had seen a dissimilar body-sized model with a BMI below average (M = 2.93, SD = 1.27), a similar body-sized model (M = 3.33, SD = 1.08), and a dissimilar body-sized model with a BMI above average (M= 3.32, SD = 1.23).
The main effect of attitudes towards the product in the ad for actual similarity in body size yielded an F-ratio of F(2,154) = 1.20, p = .30, also indicating no significant differences between the groups who had seen a dissimilar body-sized model with a BMI below average (M = 3,39, SD = 1.31), a similar body-sized model (M = 3.59, SD = 1.02), and the dissimilar body-sized model with a BMI above average (M= 3.75, SD = 1.19).
The main effect of attitudes towards the model in the ad for actual similarity in body size yielded an F-ratio of F(2,154) = .59, p = .55, again indicating no significant differences between the groups who had seen a dissimilar body-sized model with a BMI below average
36 (M = 4.25, SD = 1.32), a similar body-sized model (M = 4.49, SD = 1.01), and a dissimilar body-sized model with a BMI above average (M= 4.46, SD = 1.31).
The main effect of purchase intention for the product in the ad considering actual similarity in body size yielded an F-ratio of F(2,154) = 1.34, p = .26, indicating no significant differences between the groups who had seen a dissimilar body-sized model with a BMI below average (M = 2.61, SD = 1.29), a similar body-sized model (M = 3.04, SD = 1.24), and a dissimilar body-sized model with a BMI above average (M= 2.91, SD = 1.32).
As a result of the outcomes presented in Table 7, hypotheses 2a, 2b, 2c, and 2d are rejected.
Table 7: Summary of statistics body size similarity
Independent Variable Dependent Variable F p
Body size Overall score 1.61 .20
Attitude towards the ad in general 1.83 .16
Attitude towards the product in the ad 1.20 .30
Attitude towards the model in the ad .59 .55
Purchase intention 1.34 .26
Hypothesis 3: Interaction effect between skin color and body size
The third hypothesis states that there is an interaction between similar skin color and similar body size, via skin color that increases the positive effects on the dependent variables. Therefore, both skin color and body size are combined in ANOVA-analyses.
Results indicated significant findings for some of the dependent variables.
37 Interaction effect attitude towards the ad in general
An ANOVA-analysis with the interaction between body size and skin color, for the dependent variable attitude towards the ad in general, presented an F-score of F(5,151) = 1.52, p = .22, indicating no significant differences. In contrast to what was expected, no significant differences are found, and therefore, H3a is rejected.
Interaction effect attitude towards the product in the ad
An F-score of F(5,151) = 2.98, p = .05 indicated a significant interaction effect for attitude towards the product in the ad, combining skin color and body size. Moreover, the univariate outcomes explained that a significant difference only existed within the similar skin-colored conditions with an F-score of F(2,503) = 3.71, p = .03. With a pairwise comparison, a significant difference was found between the condition with a dissimilar body size with a BMI below average (M = 2.96, SD = .91) and the similar body size condition (M = 3.81, SD = .87) at p = .02. Also, a significant difference was found between the dissimilar condition with a BMI below average (M = 2.96, SD = .91) and the dissimilar body size with a BMI above average (M = 3.80, SD = 1.14) at p = .02. As a result, H3b is accepted for the conditions described above. The plot can be found in Figure 3.
Figure 3: Interaction effect of body size and skin color on attitude towards the product 2,9
3 3,1 3,2 3,3 3,4 3,5 3,6 3,7 3,8 3,9
Mean score (scale 7)
Skin color model similar skin color white
dissimilar skin color black
Dissimilar body size
below average BMI Similar body size
average BMI Dissimilar body size above average BMI
38 Interaction effect attitude towards the model in the ad
A significant interaction effect was found for attitude towards the model in the ad, combining skin color and body size with an F-score of F(5,151) = 4.87, p = .01. Moreover, the univariate outcomes explained that a significant difference only exists within the similar skin-colored conditions with an F-score of F(2,665) = 4.70, p = .01. Specifically, a significant difference is found between the dissimilar body size condition with a BMI below average (M = 3.64, SD = .94) and the similar body size condition with the average BMI (M
= 4.77, SD = .88) at a p-value of .00.
Also, a significant difference was found between the dissimilar body size condition with a BMI above average (M = 4.33, SD = 1.15) and the dissimilar condition with a BMI below average (M = 3.64, SD = .94) at a p-value of .06. As a result of the significant findings, H3c is accepted for the significant differences within the similar skin-colored category between the conditions described above. The plot of the interaction effect can be found in Figure 4.
Figure 4: Interaction effect of body size and skin color on attitude towards the model in the ad
3,5 3,7 3,9 4,1 4,3 4,5 4,7 4,9
Mean score (scale 7)
Skin color model similar skin color white
dissimilar skin color black
Dissimilar body size
below average BMI Similar body size
average BMI Dissimilar body size above average BMI
39 Interaction effect purchase intention
An ANOVA-analysis with the interaction between body size and skin color, for the dependent variable purchase intention, presented an F-score of F(5,151) = .58 p = .56, indicating no significant findings. In contrast to what was expected, no significant differences are found, and therefore, H3d is rejected. In Table 8, a summary of the statistics regarding the interaction effect can be found.
In conclusion to the hypotheses related to the interaction effects between skin color and body size: H3a is rejected, H3b is partly accepted, H3c is partly accepted, and H4d is rejected.
Table 8: Summary of statistics interaction effects
Hypothesis 4: moderator ethnic identity for skin color
Ethnic identity is investigated in this study as a possible moderator between skin color and the dependent variables. An ANOVA-analysis was performed, with skin color and ethnic identity as fixed factors, to find possible significant findings. For this analysis, ethnic identity was recoded from a 7-point Likert scale into three groups. These three groups are equally divided into three categories (group 1 = 47 respondents, group 2 = 55 respondents, and group 3 = 55 respondents):
Independent Variable Dependent Variable F p
Body size * Skin color Overall score 2.92 .06
Attitude towards the ad in general 1.52 .22
Attitude towards the product in the ad 2.98 .05
Attitude towards the model in the ad 4.87 .01
Purchase intention .58 .56
